On Theft
The same article you link to says that what it calls 'theft' is the practice of multinationals dodging taxes. The article explains it the two paragraphs before:
Multinational companies also steal money from developing countries through “same-invoice faking”, shifting profits illegally between their own subsidiaries by mutually faking trade invoice prices on both sides. For example, a subsidiary in Nigeria might dodge local taxes by shifting money to a related subsidiary in the British Virgin Islands, where the tax rate is effectively zero and where stolen funds can’t be traced.
Tax dodging is not theft according to dictionary definition but I believe the author is using the word theft metaphorically here (e.g. like when Nozick when he said 'taxation is theft' or when Proudhon said 'property is theft', of course its not meant literally).
On Impact of Tax Dodging on Economic Development
This is topic that is not yet well understood as some other areas of economics due to lack of data. However, what research there is suggest that tax avoidance negatively impacts economic development (e.g. see reviews of literature such as: Fuest & Riedel, 2009; Cobham, 2005).
The reason why tax avoidance harms economic development is that it reduces the government revenue that could be in principle used for provision of public goods. Public goods such as for example, provision of effective police that protects people and their property, or effective courts etc. In turn these public goods and institutions they protect are crucial to economic development (e.g. see Acemoglu and Robinson 2012). However, what is less clear is whether some developing countries would actually in a contrafactual world where they get this extra tax revenue spent it on such public goods or used it to prop up corrupt regime (e.g. how oil revenue is often used by dictatorships to stay in power).
However, the caveat above being said, its not unreasonable to state that such tax dodging probably negatively affects economic development. Even if the evidence is not completely conclusive its thought that it is more likely than not to have negative effect.